- index.ts: registerTool() does the session check, try/catch and the
write queue for all 15 tools; fail()/ok(), syncFromBrowser(),
commitDocument() and describePages() replace the copied blocks; drop
pickPageSelector, currentSession.version, the input ?? {} defaults;
add_page reuses prepareNewModel() split out of prepareNewDiagram
- xml-validation.ts: drop the hand-written syntax checks that saxes
already reports, MAX_XML_SIZE and the stale header
- http-server.ts: one pendingExport object, no unread response fields,
no CORS/OPTIONS branch for same-origin requests, fewer exports
- diagram-operations parseNewCell(), addHistory returns void,
normalizeToMxfile without host, foldCells without layer, expandStyles
removed, unused preview.css rule removed, no .d.ts output
- xml-examples.ts styleRules() shares the Styles paragraphs with the
web prompt
- tests: setupFiles installs the DOM polyfill once, server-wiring uses
the SDK client, mcp-core.test.ts drops the cases MCP tests already
cover
The model defines reused styles once as <mxStyle name="..." value="..."/> and
refers to them by name, and writes shapes as one self-closing mxCell with
x, y, w, h and edges with source and target. style-classes.ts and
compact-cells.ts expand both back into standard draw.io XML after validation
and add draw.io's html=1 / whiteSpace=wrap defaults; the diagram shown to the
model is folded into the same notation. Prompts, tool descriptions and the MCP
drawing guide teach the notation with shared examples.
Measured on the five start-screen examples against main: gpt-6-luna 35% fewer
output tokens, Claude Opus 5.5 41% fewer. Two five-model review rounds fixed
quote-aware cell matching, attribute escaping, edge/vertex inference, root-id
and layer handling, and several auto-fix gaps.
dayuan.jiang